Abstract: In order to realize the static identification of mine external fire, the calculation methods of roundness and rectangularity of image contour were investigated and the detailed implementation scheme of sharp angle feature identification under external fire identification were given. Considering the complex mine camera system had the grey characteristics of unascertained structure, incomplete parameters and integral effect on imaging error, GM(1, 1) grey model and metabolic iterative modeling method were used to predict the evolution trend of measurement error. The distance between the flame and the camera was used as the original data to build the gray model, and the latest depth-of-field data metabolism was used for iterative optimization. The results show that the large baseline camera can reduce the distance measurement error, and the grey modeling method based on machine vision and metabolic mechanism can effectively improve the accuracy of external fire identification and positioning. When the development coefficient is less than 0.3, the prediction accuracy within five steps is more than 97%, which is suitable for medium and long term prediction.